Brain scans reveal two distinct autism subtypes with different biological origins

For decades, clinicians and researchers have watched autism present itself in wildly different ways from one person to the next—different strengths, different struggles, different trajectories. The variability has been so pronounced that some have questioned whether autism is even a single condition or a collection of distinct disorders wearing the same name. But without a way to look inside the brain and see what was actually different at a biological level, that question remained mostly unanswered. Now, an international team of neuroscientists has found evidence that at least two fundamentally different types of autism exist, each rooted in separate biological mechanisms.

The research, published in Nature Neuroscience and led by Alessandro Gozzi at the Italian Institute of Technology in Rovereto and Adriana Di Martino at the Child Mind Institute in New York, took an unusual approach: the team used mouse models to decode what was happening at the molecular level, then used that knowledge as a guide to identify matching patterns in human brain scans. They analyzed functional connectivity in twenty different mouse models, then compared those findings against fMRI scans from 940 children and young adults with autism, alongside scans from more than 1,000 neurotypical individuals. What emerged were two reproducible subtypes, each with its own signature pattern of brain communication.

In one subtype, which the researchers call hyperconnectivity, brain regions communicate more intensely with one another than they do in typical development. In the other, hypoconnectivity, communication between brain areas is diminished. The mouse work revealed that these patterns were not random variations—they were tied to specific biological pathways. The hypoconnectivity pattern correlated with alterations in synaptic function, the mechanisms by which neurons pass signals to one another. The hyperconnectivity pattern, by contrast, was associated with immune-related systems in the brain. When the team looked at human brain tissue, they found the same molecular signatures: regions showing reduced connectivity were enriched for genes involved in synaptic transmission, while hyperconnected regions showed enrichment for immune-related genes.

Together, these two subtypes accounted for roughly a quarter of the autistic individuals in the study. The hyperconnectivity subtype showed modestly higher scores on standardized autism severity measures, suggesting that the two groups might experience autism somewhat differently in their daily lives. Yet the most striking finding was simply that the subtypes held up. When the researchers tested their patterns against independent datasets from dozens of research sites worldwide—data aggregated through the Autism Brain Imaging Data Exchange, a neuroimaging initiative co-founded by Di Martino—the same two subtypes emerged. That reproducibility across different laboratories, different populations, and different scanning protocols is the kind of validation that transforms a hypothesis into something closer to biological fact.

Gozzi described the mouse models as a kind of biological "Rosetta Stone," a way to translate molecular mechanisms into the brain imaging patterns visible on an fMRI scan. The team could identify which genetic and immune factors drove which connectivity signatures in mice, then search for those same signatures in human brains. This bridging between animal models and human neurobiology is what allowed them to move beyond simply observing that autistic brains look different—they could now say something about why.

The implications point toward precision medicine. If autism comes in at least two biologically distinct forms, then treatments and interventions might need to be tailored accordingly. A therapy designed to address synaptic dysfunction might not help someone whose autism stems from immune dysregulation, and vice versa. The current findings offer a foundation for developing such targeted approaches, though the researchers are careful to note that the full spectrum of autism likely encompasses additional subtypes beyond these two. Larger datasets and more refined analytical methods will probably reveal further diversity. For now, though, the work demonstrates that the variability clinicians have long observed in autism has a biological basis—and that basis can be measured, reproduced, and potentially used to guide care.